Detection of Money Laundering Using Graph Neural Networks and Transformer-Based Learning

Abstract

Money laundering is one of the most significant financial crimes affecting banking institutions, governments, and global financial systems. Criminal organizations use sophisticated transaction strategies to conceal the origins of illegally obtained money and integrate it into legitimate financial systems. Traditional Anti-Money Laundering (AML) systems rely heavily on rule-based monitoring approaches that often fail to detect evolving laundering patterns and generate excessive false positive alerts.This research proposes a hybrid Graph Neural Network (GNN) and Transformer-based deep learning framework for intelligent money laundering detection. The proposed system models financial transactions as graph structures where customer accounts represent nodes and transactions represent edges. Transformer-based attention mechanisms are integrated for sequential transaction analysis and contextual behavior learning. Experimental evaluation demonstrates that the proposed framework achieves superior detection accuracy, higher recall, improved precision, and lower false positive rates compared to conventional machine learning and deep learning techniques. The framework is scalable and suitable for deployment in next-generation AI-driven AML systems.